The f(x)=e^x | Model collapse and ironies in the world of AI

The f(x)=e^x | Model collapse and ironies in the world of AI
👋 Hi, I am Mark. I am a strategic futurist and innovation keynote speaker. I advise governments and enterprises on emerging technologies such as AI or the metaverse. My subscribers receive a free weekly newsletter on cutting-edge technology.

ChatGPT's joke of the week:

Why was the AI model kicked out of school?

Because it was caught letting its "ChatGPT buddy" do all its homework!

The Danger of AI Model Collapse: When LLMs are Trained on Synthetic Content

My latest article:

Navigating the world of AI can be a tightrope walk, especially when faced with the phenomenon of model collapse. This occurs when LLMs become overly reliant on data created by the LLM, something that could happen with future iterations of existing large language models as the internet is now flooded with synthetic content created by those LLMs.

Model collapse in AI can lead to distorted perceptions of reality and perpetuate biases. This can compromise the reliability and accuracy of AI models, affecting their ability to provide accurate predictions or recommendations.

Potential solutions include verifying synthetic content before using it as training data and creating language models using smaller datasets. The effectiveness of these strategies varies based on factors like model architecture, dataset composition, and training objectives.

You can visit to book me as a keynote speaker for your next event. Don't forget to read my top ten technology trends for 2023 if you want to know what is in store for this year, or watch my speaker demo:


Useful Nuggets of Information

My weekly tips from around the web to get you thinking.

1. AI's dark side: junk websites drain ad budgets.

AI-generated junk websites are siphoning off online ad money from major brands, leading to billions wasted annually. As AI advances, this worrying trend is set to escalate. (MIT)

2. ChatGPT's new updates: power and pitfalls.

OpenAI is set to roll out significant updates for ChatGPT, including a new function-calling feature that allows it to interact with the outside world. While this enhances its utility, it also raises concerns as ChatGPT cannot distinguish fact from fiction. (Mashable)

3. Meet Voicebox: Meta's ai that mirrors your voice!

Meta's new AI, Voicebox, can mimic voices with stunning accuracy. This powerful tool can generate natural audio clips and even replicate the style of a given voice sample, marking a significant advancement in AI technology. (ZDNET)

4. Humans paid to train AI, outsource the job... to AI.

In an ironic twist, many people paid to train AI models are actually outsourcing their work to AI itself. This ironic cycle could lead to more errors in AI models, creating a self-defeating loop in the AI world. (MIT)

Trees getting axed while pledges grow: the sad irony of deforestation.

Despite lofty promises to halt deforestation, forests continue to fall. It's a tragic irony: we pledge to save trees while simultaneously cutting them down, risking ecosystems and our fight against climate change. (Vox)

Know someone who needs the f(x) = e^x?

Forward it to someone who might like it too.

Subscribe here if it was forwarded to you.


Digital Twin